Upload
volien
View
224
Download
1
Embed Size (px)
Citation preview
Atmospheric Composition Modelling and Assimilation
(with focus on aerosols)
Angela Benedetti With contributions from:
Alessio Bozzo, Francesca Di Giuseppe, Luke Jones, Samuel Rémy, Mark Parrington, Anna-Augusti Panareda, Vincent-Henri Peuch,
Sebastien Massart, Antje Inness, Johannes Flemming, Miha Razinger, Martin Suttie, and the CAMS team
OUTLINE • General introduction to the Copernicus Atmosphere Monitoring
System (CAMS)
• Data products and catalogue
• Interesting cases: the Indonesian Fire season of 2015
• Overview of modelling and data assimilation efforts with focus on aerosols
• Impact of aerosols on NWP (medium-range and long-range)
• Summary and future perspectives in aerosol prediction
THE COPERNICUS ATMOSPHERE
MONITORING SYSTEM
(CAMS)
Atmospheric composition is a pivotal element between human activities and the Earth Environment
Atmospheric composition and its changes affect our health and well-being
climate change
emissions
exposure
impacts
ozone
NOx greenhouse gases
aerosol
PM
adaptation
mitigation
Why?
CAMS: A Significant Heritage • A decade-long series of R&D projects and an internationally respected European achievement (GEMS, MACC, -II, -III) • An equally long experience in engaging with users and potential users in Europe and across the world (PROMOTE, MACC, -II, -III)
Strategy Socio-economic impact Experts Users
From Earth Observation to policy-quality products
2010
2011
Over 70 EO instruments are assimilated in the global system
Policy-relevant (here health indicator for ozone) products are delivered. They are “maps with no gaps”, which observations alone don’t provide and are essential to assess impacts.
Boundary conditions feed an ensemble of high-resolution European AQ systems (in
order to assess uncertainties)
More data are assimilated (in particular in situ) and used for extensive validation
© ECMWF
AIR QUALITY AND ATMOSPHERIC COMPOSITION
European air quality analyses, forecasts and assessments in support of reporting and policy making, pollen forecasts, global transport of constituents/pollutants.
CLIMATE FORCING
Distributions of aerosol components and their radiative impacts, other radiative forcings.
OZONE LAYER AND UV
Monitoring and forecasting of the ozone layer / hole, UV index, UV radiation (crops, ecosystems).
SOLAR RADIATION
Estimates of solar irradiance at surface, improved potential yield assessments for solar plants.
EMISSIONS AND SURFACE FLUXES
Estimates of human emissions globally and in Europe (high-resolution), emissions by wildfires, surface fluxes of CO2, CH4 and N2O.
CAMS Portfolio
http://www.copernicus-atmosphere.eu
Search criteria based on service themes, species,
geographic area, etc.
Products found
Pop-up window with product description and links to plots,
data, and validation
CAMS online catalogue search
(open data policy)
http://www.copernicus-atmosphere.eu
NRT / on-line evaluation
NO2, Europe-wide, ~15 km, hourly +96h
Multi-model spread as a measure of forecast uncertainty
FORECAST PRODUCTS
Global and European maps of major pollutants
Daily time-critical users of Global Services
Daily time-critical users of Regional Services
Users of the global re-analysis
atmosphere.copernicus.eu web
Islandic volcano event
GROWING CAMS AUDIENCES (3000+ USERS)
RECENT EPISODES Poor air quality over Western Europe (March 2015 )
PARIS
Dust advection from the Sahara (March 2015)
Indonesian fires –large biomass burning AOD anomaly (Aug-Sep-Oct 2015)
INDONESIAN FIRES (AUG-OCT 2015)
Fire Radiative Power (W/m2) accumulated over Indonesia during the 2015 fire season (Aug-Oct). Credits: Francesca Di Giuseppe
INDONESIAN FIRES (AUG-OCT 2015)
Biomass burning AOD anomaly: up to 2000%
CO anomaly: up to 500%
O3 anomaly: 30-40 %
Benedetti et al, 2106 to appear in State of Climate, BAMS. Credits: Antje Inness, Mark Parrington (ECMWF), Gerry Ziemke (NASA)
AEROSOL MODELLING AND
ASSIMILATION
CAMS aerosol forecasts
• Built on the ECMWF NWP system with additional prognostic aerosol variables (sea salt, desert dust, organic matter, black carbon, sulphates)
• Aerosol data used as input in the aerosol analysis: - NASA/MODIS Terra and Aqua Aerosol Optical
Depth at 550 nm - EUMETSAT’s PMAP AOD (monitoring) - NASA/CALIOP CALIPSO Aerosol Backscatter
(experimental) - AATSR, SEVIRI, VIIRS (experimental) • Verification based on AERONET Aerosol Optical
Depth (and now also Angstrom exponent) • Part of multi-model ensemble efforts such as the
International Cooperative for Aerosol Prediction (ICAP) and the WMO Sand and Dust Storm Warning and Assessment System (SDS-WAS) North-African-Middle-East-Europe and Asian nodes.
Source: WMO SDS-WAS http://sds-was.aemet.es
Source: ICAP http://icap.atmos.und.edu/
Aerosols in the ECMWF IFS (C-IFS)
Morcrette et al. 2009, JGR, 114, doi:10.1029/2008JD011235
12 aerosol-related prognostic variables: * 3 bins of sea-salt (0.03 – 0.5 – 0.9 – 20 µm) * 3 bins of dust (0.03 – 0.55 – 0.9 – 20 µm) * Black carbon (hydrophilic and –phobic) * Organic carbon (hydrophilic and –phobic) * SO2 -> SO4 Physical processes include: • emission sources (some of which updated in NRT, i.e.fires), • horizontal and vertical advection by dynamics • vertical advection by vertical diffusion and convection • aerosol specific parameterizations for dry deposition, sedimentation, wet deposition by large-scale and convective precipitation, and hygroscopicity (SS, OM, BC, SU)
Recent developments: Dust emissions • Overestimation of dust AOD : the
average among the models participating in AEROCOMS is 0.023
• Compared to the literature and other models, the amount of larger particles in dust emissions is too low.
• => decrease of the amount of small particles in the emissions, increase the amount of larger particles
Global dust AOD for May 2014 as a function of lead time, with (red) and without (blue) data assimilation
Credits: Samuel Rémy
• Better balance between the model and observations after the introduction of new emissions
AOD at the AERONET station of Tamanrasset (Algeria), from 15/4/2014 to 1/8/2014. Observations (blue), old emissions (red) and new emissions (black)
Recent developments: Injection heights for biomass burning aerosol emissions
• Biomass burning emissions are currently emitted at the surface. • Injection heights for biomass burning emissions are routinely produced by
GFASv1.2., using a Plume Rise Model (Freitas et al, 2007, Paugam et al., 2015), and Sofiev’s parameterization (Sofiev et al. 2012)
• Use of these injection heights was implemented in CIFS for aerosols, chemical species, greenhouse gases
Profile of OM mixing ratio over Canada (52N, 77.5W) on July 6, 2013 Blue, emissions of OM at surface, red, emissions at the injection height given by the PRM
Credits: Samuel Rémy
Evaluating the impacts of smoke injection heights computed from plume rise model
• Injection heights for smoke emissions are estimated using a Plume rise model (Paugam et al., 2015, based on Freitas et al., 2007)
• This plume rise model uses MODIS FRP and modelled atmospheric profiles with a shallow convection scheme to represent detrainment from fire plumes
• Initial comparisons show that both aerosol extinction and AOT increase throughout the profile, not necessarily at smoke height shown in DIAL/HSRL profile
DIAL/HSRL
MACC-III with plume rise model
MACC-III
August 19 August 27 DIAL/HSRL
MACC-III with plume rise model
MACC-III
Credits: Rich Ferrare and Sharon Burton (NASA Langley)
Evaluating the impact of higher model resolution
• Model resolution increased from T255 (80 km) with 60 vertical levels to T1279 (16 km) with 137 vertical levels
• Higher resolution represents smoke altitude better than assimilating MODIS AOT or using plume rise model
DIAL/HSRL
MACC-III T1279 (137 levels)
MACC-III T255 (60 levels)
DIAL/HSRL
MACC-III T1279 (137 levels)
MACC-III T255 (60 levels)
August 19 August 27
Credits: Rich Ferrare and Sharon Burton (NASA Langley)
Future: GLOMAP aerosol in C-IFS
OH
OH, NO3
DMS
SO4 OC BC
NH4 NO3
Aitken accum
SO2
H2SO4
SEC_ORG
MONOTER
coarse
SO4 OC NH4 NO3
SO4 OC BC Na DU
NH4 NO3 Cl
SO4 OC BC Na DU
NH4 NO3 Cl
soluble Soluble nucln soluble soluble
component mass conc.
N1 N2 N3 N4
OH, NO3, O3 Sulphate, sodium, black carbon, organic carbon, dust Ammonium, nitrate, chloride
BC OC
Insoluble Aitken
N5
accum coarse insoluble insoluble
DU DU
N6 N7
Aerosol mass as “components” in internally mixed modes
29 mass 7 number
Transported tracers=36
Also emit sea-spray as sodium chloride & sodium sulphate. Account for sea-salt-SO4 as well as usual nss-SO4.
NH3
HNO3
HCl
Credits: Graham Mann (University of Leeds)
Evaluation suite for assessing IFS- GLOMAP (also in UM, TOMCAT)
• GLOMAP evaluation strategy involves assessing a range of aerosol metric against observations. As well as aerosol optical depth speciated mass, size-resolved number concentrations are used.
Sulphate mass evaluation against EMEP, IMPROVE, U. Miami obs datasets for reference IFS-GLOMAP run
Credits: Graham Mann, Sandip Dhomse (Uni Leeds)
The ECMWF 4D-Var
The observations are used to correct errors in the short forecast from the previous analysis time. This is done by a careful 4-dimensional interpolation in space and time of the available observations.
Every 12 hours we assimilate 4 – 8,000,000 observations to correct the initial conditions on the 100,000,000 variables that define the model’s virtual atmosphere (winds, temperature, humidity, surface pressure, ozone and surface variables for the standard operational configuration).
Additional variables are included in the control vector for the MACC NRT analysis and forecast (reactive gases and aerosols).
The aerosol analysis • Integrated in the ECMWF incremental 4D-Var
• Control variable is formulated in terms of the total aerosol mixing ratio.
• Increments in total mass are repartitioned into the single species according to their
fractional contribution to the total.
• Background error statistics have been computed using forecasts errors as in the NMC method (48h-24h forecast differences).
• Assimilated observations are the MODIS Aerosol Optical Depths (AODs) at 550 nm over land and ocean, including Deep Blue over bright surfaces. Observation errors are prescribed fixed values.
• A global variational bias correction with constant and surface wind predictors for MODIS data is implemented in the current near-real time run.
Benedetti et al. 2009, JGR ,114, doi:10.1029/2008JD011115
Aerosol Optical Depth coverage from various sensors/products
AATSR: data over deserts but narrow swath & one Instrument. Can be replaced by SLSTR on Sentinel-3
PMAP: for now, only data over ocean were tested at ECMWF. Two platforms (more resilient), multi-sensor (more points of failure).
MODIS: two platforms, global coverage. Ageing. Data also over bright surfaces when Deep Blue is used. SEVIRI: geo-stationary, high
data volume, partial coverage
AATSR Aerosol Optical Depth data AATSR data from FMI were used in a special Climate Change Initiative reanalysis for 2008 • Adds value to forecast-only run as shown by comparison with AERONET data • Less impact than MODIS in the analysis due to coverage • Data from Sentinel 3 SLSTR are expected to have a similar impact
Forecast-only run AATSR-only run MODIS-only run MODIS and AATSR run
• Other experiments also showed a positive impact of the AATSR data
PMAP Aerosol Optical Depth
Forecast-only run PMAP-only run MODIS-only run
Produced pre-operationally by EUMETSAT based on GOME2, AVHRR and IASI data. Similarly to AATSR data: • Adds value to forecast-only run as shown by comparison with AERONET data • Impact comparable to MODIS due to global coverage
• Monitoring of PMAP has started recently • Assimilation will follow
SEVIRI Aerosol Optical Depth (ocean-only)
SEVIRI + MODIS run MODIS-only run
• Produced in NRT at ICARE http://www.icare.univ-lille1.fr/msg/ • Based on an algorithm by Thieuleux et al., 2005 • Small but detectable impact on global bias (negligible in RMS) • European/African coverage • Of interest for European regional data assimilation • Huge data volume (thinning needed) • Other products under consideration
Data coverage over 24h
Assimilation of lidar signal
Data: all operational data plus MODIS AOD and CALIOP Level 1.5 backscatter
Lidar backscatter x 1e7 (sr m)-1
• Expedited product (courtesy of CALIPSO team at NASA Langley: David Winker, Chip Trepte, Jason Tackett)
• Average attenuated backscatter at 20 km, cloud-
cleared at 1 km. • 345 vertical levels corresponding to 60 m resolution (averaged to 300 m before assimilation)
CALIOP level 1.5 sample orbit August 18, 2010
Verification of lidar assimilation experiments
CALIOP + MODIS (both bias corrected)
MODIS only CALIOP+ MODIS
…but globally the MODIS-only run is still slightly on the lead.
AERONET verification shows good performance of lidar assimilation locally or at least not worse than the MODIS Dark Target-only run….
Evaluation of the impacts of CALIOP profile assimilation
• Assimilation of CALIOP profiles slightly reduces extinction profiles in some locations; largest extinction values remain near surface
• Depending on location, these reductions can improve or worsen agreement with HSRL
DIAL/HSRL DIAL/HSRL
MACC-II
August 19 August 27
MACC-III with MODIS AOT assimilation
MACC-III with MODIS AOT assimilation
MACC-III with MODIS AOT assimilation and CALIOP assimilation
MACC-III with MODIS AOT and CALIOP assimilation
Credits: Rich Ferrare and Sharon Burton (NASA Langley)
Comparison of Median Profiles with and without CALIOP assimilation
• Median profiles in good
agreement with MODIS AOT assimilation
• Adding CALIOP: – produces relatively minor
effects on median profiles – tends to lower the AOT with
respect to runs that assimilate only MODIS AOT
– gives a slightly better agreement with HSRL
MODIS assimilation only
MODIS and CALIOP assimilation
HSRL MACC-III
HSRL MACC-III
HSRL MACC-III
HSRL MACC-III
Credits: Rich Ferrare and Sharon Burton (NASA Langley)
CAMS REANALYSIS RUNS • New “interim” reanalysis from 2003-2015 has been run in parallel mode (literally)
for fast turnaround • Overall good performance • Used for contribution to the State of Climate (BAMS, in publication) • Substantial differences with MACC reanalysis
MACC (MODIS) MACC reanalysis
Interim reanalysis
Flemming et al 2016, APCD
REANALYSIS RUNS: BAMS STATE OF CLIMATE 2015
TOTAL AOD TREND 2003-2014
Rémy et al, 2016: [Global climate] Aerosols [in "State of the Climate in 2015"]. To appear in Bull. Amer. Meteor. Soc.
TOTAL AOD 2003-2014
AOD ANOMALY 2015
AEROSOL IMPACTS ON
NUMERICAL
WEATHER PREDICTION
Climatological AOD 550nm distribution MACC vs Tegen et al 1997 (OPER)
• MACC run (2003-2012): sources of biomass burning from GFAS, sulphate aerosol precursor from EDGAR 4.1, prognostic for sea salt and dust, revised dust model
• Optical properties recomputed for RRTM spectral bands and for each aerosol type/size bin. Mass mixing ratio as input to radiation
• Vertical distribution following an exponential decay with scale height derived from the MACC model for each aerosol type. Monthly varying for dust.
Credits: Alessio Bozzo
U 925 hPa – change in model bias
U 925 hPa – model bias at D+5
June-July
Impacts on FC errors
Credits: Alessio Bozzo, Linus Magnusson
20 year run
OLD climatology NEW climatology GPCP (obs)
WMO Working Group on Numerical Experimentation
(WGNE) This inter-comparison aims to evaluate the impact of aerosols on Numerical Weather Prediction Three situations were proposed : • Dust storm over Egypt on 18th of April 2012 • Extreme pollution over Beijing, 12-16th of January
2013 • Extreme biomass burning over Brazil in
September 2012 during the SAMBBA field campaign
Participants : Météo-France, Met-Office, JMA, ECMWF, NOAA, NASA, CPTEC (Brazil)
MODIS imagery, 18/4/2012
Beijing , 14/1/2013
Credits: Samuel Rémy
Dust case of April 2012 – Impact on temperature, winds and dust production
Rémy et al, 2015, ACP, doi:10.5194/acp-15-12909-2015
2m temp
10-m winds
Dust production
Difference between run with interactive aerosols (TOTAL_ASSIM) and reference run (REF_ASSIM) 36 hour forecast (valid on April 18th at 12UTC) • Reduced 2m temperature • Increased surface winds • Increased dust production
Aerosol impacts on monthly forecasts (I)
In collaboration with: Fréderic Vitart (ECMWF)
2003-2014 2003-2014
• Preliminary results confirm the positive impact (reduction in bias) of the interactive aerosols on meteorological fields (winds and precipitation)
• More prominent (positive) impact over the Indian Ocean and to a lesser extent in other areas • Aerosol fields will be evaluated too by comparing with the MACC/CAMS reanalysis (BONUS:
aerosol seasonal prediction!)
CONTROL RUN – WIND BIAS WEEK 4 INTERACTIVE AEROSOL RUN – WIND BIAS WEEK 4
Aerosol impacts on monthly forecasts (II) • Scorecards measures performance of interactive aerosol experiment with respect to the control run for several parameters. • Green circles indicate positive impact • Solid circles indicate significant Impact • Temperature and winds greatly improved at week 4 in the interactive aerosol run! • Upper-level temperature also improved on the seasonal scale (month 2-3)
Summary and future perspectives • CAMS offers many services related to atmospheric composition from daily forecasts to reanalysis runs both at the global and at the regional (European) level • Model developments have been carried out for the past 10 years during precursors projects. They are now part of the ECMWF’s Integrated Forecasting System • Several datasets are routinely assimilated and more are in the pipeline (Copernicus Sentinel satellites)
• The impact of interactive aerosols on Numerical Weather Prediction is being investigated at different time ranges and promises interesting results